Entanglement engineering of one-photon wavepackets using a single-atom source

Entanglement engineering of one-photon wavepackets using a single-atom source

00
On
On

X> ■
in :

(N .
(N '

> :
o .
\o ■
o ■
<n :
o .

00 ■
On 1

^— > ■

G ■

a ;

3 :

K.M. GheriW, C. Saavedra^ 1 * 2 ), P. ToiW 1 ), J. I. ChW 1 ' 3 ), and P. Zollert 1 ' 3 )
(1) Institut fur Theoretische Physik, Universitdt Innsbruck, Technikerstrasse 25/2, A-6020 Innsbruck, Austria
(2) Departamento de Fisica, Universidad de Conception, Casilla 4009, Conception, Chile
(3) ITP, University of California, Santa Barbara, CA 93106-4030
(February 1, 2008)

We propose a cavity-QED scheme for the controlled generation of sequences of entangled single-
photon wavepackets. A photon is created inside a cavity via an active medium, such as an atom, and
decays into the continuum of radiation modes outside the cavity. Subsequent wavepackets generated
in this way behave as independent logical qubits. This and the possibility of producing maximally
entangled multi-qubit states suggest many applications in quantum communication.

Pacs number(s): 03.67.Hk, 03.67.-a

Sources offering a great variety of entangled states are
required for the implementation of many quantum com-
munication and computation protocols jl],^|. With quan-
tum communication || in mind the choice of photons as
qubits is especially appropriate, since they can be eas-
ily transfered over long distances. The standard source
presently used in the lab is parametric downconvcrsion
in a crystal |^]||. It is a reliable source of entangled
twin-photons but the process is random and largely un-
tailorable. Moreover, in practice its capability of gener-
ating entanglement is limited to states comprising only
two photons. In this Letter we propose a scheme for the
controlled generation of many entangled photonic qubits.
Our source of entanglement produces a train of single-
photon wavepackets which are well resolved in time. This
permits us to regard them as individual qubits. In its
most simple implementation the setup consists of a sin-
gle multilevel atom inside an optical resonator |^^] . The
individual wavepackets are generated by applying an ex-
ternal laser pulse to the atom prepared in a superposition
state of its internal states. The coupling of the atom to
the resonator allows the transfer of a single photon to
the resonator and therefrom via cavity decay to the con-
tinuum of radiation modes outside the resonator (possi-
bly coupled to an optical fiber). An encoding of quan-
tum information in the one-photon wave-packets could
either take place by identifying two orthogonal polariza-
tion states of the single photon with logical "0" and "1" ,
or by regarding the absence of a photon as logical "0"
while its presence would correspond to logical "1".

Our scheme offers a twofold advantage over already
existing sources of entangled single-photon wavepackets
such as down-conversion. It provides excellent control
over the instances in time when a qubit is created as well
as over the spectral composition of the wavepacket. The
qubits may thus be generated with a well defined tact
frequency and pulse shape. Moreover, repeated coherent
recycling of the state of the atom after the generation of
a photon wave packet gives rise to higher order entangle-
ment between subsequent photons H . In this regard our
scheme generalizes and extends recent work on sources of
single photon wavepackets, commonly referred to as pho-

ton guns (9) or turnstile devices |l(J by allowing the gen-
eration of entangled multiphoton states. In particular,
states such as the the three-particle GHZ state, and more
generally n-qubit maximally entangled states (MES) can
be generated. Since the individual wavepackets do not
overlap in time, each can be sent to a different receiver
node using simple classical gating operations. Such high-
order entangled qubit states have immediate application
in quantum cryptography |TT| and teleportation ]l2[ ], as
well as in tests of non-locality and multiparticle inter-
ference [ [[3) . Because entangled states of more than two
qubits can be generated in a straightforward manner, our
scheme is especially useful for quantum communication
between many parties Jwj.

Whilst the theory underlying our proposal can be for-
mulated in a model-independent fashion it is more in-
structive to illustrate the basic ideas using a specific
model: we consider a single atom or ion trapped inside
a cavity fflfy. For the atom we assume a double three-
level A structure in the large detuning limit as depicted in
Fig. 0. The levels \i a ) ( a = 0,1) are coupled to the upper
levels \r a ) via classical fields Q, a {t)e~ l ^ at+ ^ a ^\ where
uj a are the field center frequencies and the subscript refers
to the two polarization states. The external control pa-
rameters are the real amplitudes fl a (t) and the phases
4> a {t). The levels |/ Q ) are coupled to the upper levels
by the cavity modes a a (common frequency lu c but or-
thogonal polarization), with coupling constants g a . The
large detuning (S) assumption allows us to adiabatically
eliminate the upper atomic levels. We are left with two
two-level systems describable by generalized spin oper-
ators Oi a j a — \i a ){ja\- The center frequencies of the
external laser pulses fulfil the Raman resonance condi-
tion. Note that any offsets can still be accommodated
in the phases 4> a (t). The field outside the resonator is
described by a continuum of harmonic oscillators with
creation and annihilation operators b^ a (uj), b a (u>), respec-
tively. The reservoir modes satisfy standard bosonic com-
mutation relations: [b a (u>),bp(v)] — 5 a p8(u) — v). The
coupling of the cavity and reservoir modes is assumed to
be fl at aro und the cavity resonance frequency and equal
to ©■

1

Jr.) j£i»>
~%) *® lie) |S)

FIG. 1. A single atom with six internal states interacts
with two cavity modes of orthogonal polarization ao, ai. In a
Raman process (step 1) an initial superposition state of lev-
els |io) and is transformed into an entangled cavity-atom
state. Due to cavity leakage the photon will leave the cavity
and produce a photon wave packet in the continuum modes
outside the resonator. In step 2 the atom is recycled back to
\io) and Between two photon generations levels |io) and
can be coupled (step 3) to tailor the outgoing state.

Hence the total system consists of three building
blocks: the continuum outside the resonator, the cavity
modes and the internal degrees of freedom of the atom
inside the resonator. We switch to an interaction picture
with respect to the free dynamics of the compound sys-
tem. This eliminates the fast optical timescales from the
dynamics and leaves us with a simpler Hamiltonian:

H(t) = £ (i^ J duo (aabiiCuy^ - H. c.)) + V(t),
V(t) =^fe(i)a iaiQ + \g a \ 2 aia a a fafa

Pa(t) (c

We have introduced the following abbreviations Q, a (t) =
Q a (t)/2-y/8, g a = g a /VS, r a (t) = g a Cl a (t) and Q =
cu + lu c . The time and intensity dependent terms in V(t)
correspond to ac-Stark shifts arising from the adiabatic
elimination of the upper atomic levels.

We assume an atom initially prepared in a superposi-
tion state \ip(0)) a = co|irj) + c iKi) which we wish to map
onto a superposition of continuum (reservoir) excitations.
The cavity and the reservoir modes are in their vacuum
states. Since the dynamics contain no polarization mix-
ing terms we may independently consider those degrees
of freedom corresponding to a single index a. We may
thus work with a smaller system and intermittently drop
the index a. The final state of the total system can be ob-
tained using the superposition principle and issuing each
partial solution with the appropriate probability ampli-
tude c a . As a starting point we will discuss briefly the
generation of single photon wavepackets entangled with
the atom. Since there can be at most a single excitation
transfered to the continuum we are led to the following
ansatz for the state of the total system:

\^{t)) = |^(t))ac|0) r + / dw|¥»o(t))ac|W-

(1)

Here |ys(i))ac and \(p(t)) ac denote atom-cavity states
with and without the transfer of a photon having taken

place into the reservoir mode with frequency Co, respec-
tively. Note that |y>(t)) ac describes the atom-cavity state
before the photon has been lost to the reservoir:

|^W>ac = a(t) e - ie W| l )|0) c + C / We-^l 2t |/)|l) c ,

where 9(t) = J* dt'fl 2 (t'). Applying a to this state
projects the coupled atom-cavity system into the state
|/)|0) c which is not coupled by V. We thus find:

|<Aa(*))ac =

dt

a\y(t')) s

(2)

If we insert this expression into the equation for | (/?(£)) ac
and perform the Markov approximation we arrive at a
simple closed equation:

-(K c a^a + iV(t))\ip(t)) a

(3)

This now permits us to specify the sought evolution equa-
tions for the coefficients in the ansatz for |y>(t)) ac :

Ci(t) = -r(t)Cf(t)exp(iO e (t)),

C f (t)= f dt'r(t')a{t')eM-(K c (t-t / )+ i6 c {t'))),
Jo

where 9 c (t) = 9(t) + <j)(t) — \g\ 2 t. In the limit of an over-
damped cavity the integral will get a non-zero contribu-
tion only from those times t' which are close to t on the
scale of the cavity lifetime k" 1 . To good approximation
it thus holds that

where fi(t) = r 2 (t)/n c , and jii(O) = 0. The actual ob-
ject of interest is the state of the continuum of radiation
modes outside the cavity. We thus insert the above result
for Cf into Eq. (0) and find:

IM*))

(4)

G(u,t)a fi \i)\Q) t

Eq. (|J) indicates a direct mapping of the initial atomic
state to the final one accompanied by the creation of a
wavepacket with spectral envelope G(u>,t), cf. step 1 in
Fig. 0. To make the scheme practical two constraints
have to be kept in mind. First of all we would like to im-
plement an efficient transfer of the photon to the contin-
uum. Secondly, we are only interested in pulse sequences
that terminate after a finite time T> k^ 1 . This would
be warranted if at the time t — T the atom is with near
certainty in its internal state |/), i.e., iff /i(T) 3> 1. Re-
calling the definition of n(i) this sets for any given pulse
duration T a lower bound for the minimum size of the
pulse area of the applied laser field. Under this assump-
tion the total system state for times t > T is given by:

2

\m) = f^ Cct i?t(o,r)|/ Q )j|o>.|o> Co |o) Cl , (5)

where B^ (tj , T) is the creation operator of a one-photon
wavepacket with logical or polarization state a within the
time window from tj to tj + T:

Bl{tj,T) = J dve^G^TpjQ). (6)

Note that the spectral envelope now carries a subscript as
the system parameters need not be the same for each of
the effective two-level systems we use to implement the
mapping. In brief, we have shown how one can trans-
form an initial atomic superposition state into an entan-
gled atom-continuum state. Had we used only a single
A system we would have recovered the photon gun || , a
tailorable simple single-photon source. The novel aspect
here is the residual entanglement between the internal
atomic state and the polarization state of the outgoing
wavepacket. We may harness this to create a sequence
of entangled one-photon wavepackets which are well re-
solved in time. Let us introduce the following abbrevi-
ation for a one-photon wavepacket with polarization a
that has been generated in the j-th generation sequence:
\a)j = Bl (t J -_ 1 ,T)|0) r (with t = 0). The state after the
first sequence in more compact form reads:

|V(i)) = (co|0)i|/o) + c 1 |l) x |/ 1 ))|0) Co |0) cl . (7)

Suppose we apply a further pulse sequence which recycles
the atom back to its initial state, i.e., \f a ) — ► |i Q ). Then
at a time t\ > T we reinitiate the same pulse sequence
that we have already used previously. It is plausible that
the wavepackets already generated have in the meantime
propagated far away from the cavity and thus cannot in-
fluence the renewed generation sequence. Going through
the same procedure again we obtain for t > t\ + T:

m)) = (co|0) 2 |0)x|/o) + ci|l) 2 |l)i|/i))|0) Co |Q) Cl . (8)

The residual entanglement with the generating system
can eventually be broken up by making a measurement
of the internal atomic state in an appropriate basis, e.g.,
|/o) ± For the state in Eq. (||) the resulting reservoir
state would be one of two states Co|0)i|0)2 ± Ci|l)i|l)2-
Repeating the generation process n-times followed by a
final state measurement we produce an n-photon wave
packet. Note that the description of the reservoir state
in terms of products of one-photon wavepackets implies
that the wavepackets can be regarded as independent
quantum entities. It has to be emphasized that such
a description is only possible because of the vanishing
temporal overlap of the individual outgoing wavepackets.
Actually components of the reservoir state are given by
products of the operators B^ (tj , T) applied to the multi-
mode vacuum |0) r . By construction operators belonging
to different sequences, cf. Eq. (|^), however, commute to
good approximation:

[B a (t k ,T),Bl(t j ,T)} = 6 a/3 I dwe^-^\G a {u:,T)\ 2

~ SapSjk- (9)

Formally, we may thus regard each creation operator as
acting on its own vacuum state. Physically, this cor-
responds to the fact that each wavepacket is contained
within its private time window of duration T or a box of
length cT with no overlap between successively generated
packets, cf. Fig. [|. We have numerically checked the fac-
torization assumption for a two-photon wavepacket mod-
eling the reservoir by a discrete set of 1024 "continuum" -
modes (amounting to more than 10 6 reservoir states)
embedded in a frequency window of width 40 k c . We
found that both the Markovian approximation used in
Eq. ([}]) and the factorization assumption for the two-
photon spectral density are excellent with relative errors
of the order of 10~ 3 .

For any source of entanglement it is essential to
fathom what the accessible class of states is. In gen-
eral, a state in a basis spanned by n-qubits is defined
by N = 2 n+1 — 2 independent coefficients. In our spe-
cific model the states can be tailored by coupling the
levels \f a ) by a microwave/Raman pulse inbetween the
qubit generation sequences. For example, this trans-
forms Eq. (|) into |V>) = Mo|0)i|/ o ) + Co4|0)i|/i) +
ad* 1 1) x I fx) - ci d\ 1 1) 1 1 /o)) jO) c 10) C1 • The coefficients d a
can be chosen at will. With each applied pulse two in-
dependent parameters are introduced. For n qubits we
have thus 2n free parameters at hand for the purpose
of state engineering. Since this is much less than N,
only a restricted subclass of states can be created. How-
ever, we emphasize that the accessible class of states in-
cludes many useful and interesting states. For exam-
ple MES such as the four Bell states, the GHZ-state
(| 000) + I 111))/ -\/2 and its higher dimensional counter-
parts [(\SX, S 2 , S n ) + |1 - Si, 1 - S 2 ,...,l - S n )) / V2,

Si = 0, 1] can easily be produced.

The maximum number n of entangled photon
wavepackets (qubits) our scheme can generate is limited
by decoherence. Relevant sources of decoherence are:
(i) laser phase and amplitude fluctuations; (ii) sponta-
neous emission during the atomic transfer; (iii) absorp-
tion in the cavity mirrors; (iv) atomic motion. The cho-
sen configuration minimizes these effects. Stabilization of
laser phase fluctuations below 1kHz represents a techni-
cal challenge. In the present scheme, the state produced
after each cycle only depends on the phase difference be-
tween the laser beams driving both transitions in Fig.
1. When these two laser beams are derived from the
same source the fluctuations in the phase difference are
effectively suppressed. Amplitude fluctuations cause a
distortion of the pulse form and lead to incomplete popu-
lation transfer. An estimate gives that n <C I/AI ~ 10 4 ,
where AI / 1 are the relative intensity fluctuations. Spon-
taneous emission from the auxiliary levels |r) at rate
r is quenched by choosing a large detuning \S\ which
leads to an effective decay rate r c ff = rf2 2 /4<5 2 with

3

il = max(f2 Q , g a ). For a peak frequency f^o = 55 Mhz
and a Gaussian pulse shape for the classical field Vl(t),
S = 1.5 GHz, g = 55 Mhz, and k c = 50 Mhz one-photon
pulse durations of around 10 cavity lifetimes are possible.
Recycling and reinitialization of the medium included a
conservative estimate would yield a generation rate of
around 1 MHz. For T = 5MHz the probability of spon-
taneous emission per cycle is < 10~ 3 . Photon absorption
in the cavity mirrors is an essential effect for high-Q op-
tical cavities. In general, it leads to two types of errors:
Photon absorption and concomitant destruction of the
entanglement. These errors are evaded by postselection
through discarding sequences with a number of detected
photons smaller than n. State distortion jl(| can occur-
even in the absence of loss of a photon according to:

l*>= E V*\x)
xe{o,i}"

J2 «»

x£{0,l}™

e -( K1 - K0 )Tnx(x)^ ( 10 )

where the x are binary representations of different photon
states, n\{x) is the number of ones contained in x, «o,i
are the loss rate for modes and 1, respectively. Errors as
in Eq. ( |Io|) vanish for a cavity with equal absorption rates
for both polarizations, i.e., kq ~ ki. The optimal way to
suppress fluctuations induced by the motion of the atom
is to place the atom at an antinode of both the cavity
modes and the laser beam (which have to be in standing
wave configuration) , where the effect of spatial variations
is minimum and operate in the Lamb-Dicke regime [ [L7J .
Finally, there might be systematic and random errors in
the adjustment of the laser pulses used in the recycling
reinitialization.

FIG. 2. Ensemble averaged fidelity as a function of the
number of qubits n. Curves (a),(c), and (e) assume 8 m —
and e m = 0.0125,0.025,0.1, respectively. Curves (b) and (d)
are the same as (a) and (c) with S m = 0.05.

To assess these effects we assume the following imperfect
mapping in each of the generation sequences: \f a ) — >
A a \i a ) + B a \f a ), where A a = (1 - e a ) exp(iS a ). In Fig.
2 we plot the fidelity T(n) of an n-qubit MES produced
by a source which is ideal exept that the magnitude and
phase of e a are evenly distributed over a range of [0, e m ]
and [— e m , e m ]ir, respectively, and the dephasing angle
is evenly distributed over [— S m ,S m ]Tr. Fig. 2 shows that
the process is rather robust against global dephasing (S a )
but that the correct timing of the 7r-pulses is critical [Q .
From curve (a) we gather that for errors in the 2% range

approximately 10 qubits can be created with a fidelity of
90%.

We have presented a CQED-based source for the con-
trolled generation of entangled rt-qubit states where
the individual qubits are nonoverlapping one-photon
wavepackets. Our model seems experimentally feasible
with state-of-the-art equipment and could form the ex-
perimental basis for multi-party communication in future
quantum networks. The theory presented can be easily
adapted to other implementations (e.g. quantum dots
or single atoms embedded in a host material jl(|) which
may emerge in the course of time as quantum systems
with long coherence times.

Acknowledgements This work was supported by the
Austrian Research Foundation under grant no. S06514-
TEC and the European TMR network ERB-FMRX-
CT96-0087. C.S. thanks Fundacion Andes for support.

[1]
[2]

[3:

[4
[5
[6
['

[«.
[9

[io;

[ii

[12
[13

[14
[15
[16

[17]

[18]

D. P. DiVincenzo, Science 270,255 (1995).

C. H. Bennett, Phys. Today 24 (October 1995) and refer-
ences cited; A. K. Ekert, Phys. Rev. Lett. 67, 661 (1991).
J. I. Cirac et al, Phys. Rev. Lett. 78, 3221 (1997); T.
Pellizzari, ibid. 79, 5242 (1997).

D. Bouwmeester et al, Nature 390, 575 (1997).

D. Boschi et al, Phys. Rev. Lett. 80, 1121 (1998).

Q. A. Turchette et al, Phys. Rev. Lett. 75 4710 (1995).
G. M. Meyer, H.-J. Briegel, and H. Walther, Europhys.
Lett. 35, 317 (1997).

E. Hagley et al, Phys. Rev. Lett. 79, 1 (1997); J. A. Bergou
and M. Hillary, Phys. Rev. A 55 4585 (1997).

C. K. Law and H. J. Kimble, J. Mod. Opt. 44 (1997) 2067.
S. N. Molotkov, and S. S. Nazin, JEPT Lett. 63, 687
(1996); A. Imamoglu and Y. Yamamoto, Phys. Rev. Lett.
72, 210 (1994); F. De Martini et al, ibid 76, 900 (1996).
C. H. Bennett et al, Sci Am. 267(4), 50 (1992); W. Tittel
et al, |quant-ph/970704S .

C. H. Bennett et al. , Phys. Rev. Lett. 70, 1895 (1993).

D. M. Greenberger et al, Am. J. Phys. 58, 1131 (1990);
D. A. Rice et al, Phys. Lett. A 186, 21 (1994).

S. N. Molotkov and S. S. Nazin, JEPT Lett. 62, 956 (1995).
C. W. Gardiner, Quantum Noise (Springer, Berlin, 1991).
T. Pellizzari et al, Phys. Rev. Lett 75, 3788 (1995); H.
Mabuchi and P. Zoller, ibid 76, 3108 (1996).
Here the size of the atomic wavepacket L is smaller than
the optical wavelength i) = i/i « 1 and the population
left behind in the wrong level is n 4 /4, which puts a limit
on the number of photon pulses n <C 4/?7 4 .
Assuming Gaussian-distributed error |e a | a simple analyt-
ical estimate yields the exponential decay in Fig. 2.

4